一种注意力专家嵌入的基于流量的生成模型的混合体,用于在单细胞RNA-Seq数据集中创建合成细胞
Sultan Sevgi Turgut Ögme1, Nizamettin Aydin2, Zeyneb Kurt3
1Department of Computer Engineering, Yildiz Technical University, Istanbul, Türkiye.
PLoS computational biology
|October 6, 2025
概括
基于流 (FB) 的生成模型,特别是基于流 (MOE-FB) 的专家混合模型,在单细胞RNA测序 (scRNAseq) 数据分析中表现出卓越的性能. MOE-FB准确地识别细胞类型,并为癌症研究生成生物学相关的合成数据.
科学领域:
- 计算生物学 计算生物学
- 基因组学就是基因组学.
- 生物信息学是一种生物信息学.
背景情况:
- 单细胞RNA测序 (scRNAseq) 对于理解组织中的细胞异质性至关重要,特别是在癌症研究中.
- scRNAseq数据提出了诸如高维度,稀疏性和不平衡的细胞类型分布等挑战,需要先进的计算方法.
- 生成模型,包括变异自编码器 (VAE) 和生成对抗网络 (GAN),越来越多地用于解决scRNAseq数据处理中的这些挑战.
研究的目的:
- 开发和评估新的基于流 (FB) 的生成模型,用于scRNAseq数据分析.
- 将FB模型的性能与VAE和GANs等既定生成技术进行比较.
- 为开发自动化 scRNAseq 数据分析系统提供指导.
主要方法:
- 在MAF-FB的基础上,开发了一种掩盖的亲缘自动回归转换嵌入式FB (MAF-FB) 模型和专家混合 (MOE) 的注意力机制,创建了MOE-FB模型.
- 通过使用胰腺组织,外周血液单核细胞 (PBMC) 和人类细胞图谱骨髓数据集进行大规模比较分析.
- 使用差异指标,自动细胞类型分类,差异基因表达分析和细胞与细胞相互作用推断来评估模型性能.
主要成果:
- 提出的FB模型,特别是MOE-FB,在所有测试的指标上都始终表现优于VAE,GAN,高斯斯和ACTIVA.
- 在集成的胰腺数据集上,MOE-FB在细胞类型分类方面取得了高准确性 (F1分数0.90,精度0.89,回忆0.92).
- MOE-FB生成了生物相关的合成数据,推断的细胞相互作用与原始数据非常相似 (RMSE 0.65).
结论:
- 基于流量的生成模型,特别是MOE-FB,为scRNAseq数据分析提供了有希望和有效的方法.
- MOE-FB在处理scRNAseq数据挑战方面表现出卓越的能力,改善了细胞类型识别和合成数据生成.
- 这些发现支持FB模型在推进生物发现的自动化scRNAseq数据分析系统方面的潜力.
相关概念视频
Flow Cytometry
15.7K
The development of flow cytometry techniques began in 1934 with initial attempts by Andrew Moldavan, a bacteriologist who counted the cells in a flowing capillary system. Moldavan pumped cells through a capillary tube focused under a microscope for visualization. The invention of photometry allowed the measurement of differentially-stained cells, and Louis Kamentsky developed the first multiparameter flow cytometer in 1965 to identify and count the cancer cells in cervical tissue specimens.
In...
In...
15.7K
Synthetic Biology
5.5K
Synthetic biology is an interdisciplinary science that involves using principles from disciplines such as engineering, molecular biology, cell biology, and systems biology. It involves remodeling existing organisms from nature or constructing completely new synthetic organisms for applications such as protein or enzyme production, bioremediation, value-added macromolecule production, and the addition of desirable traits to crops, to name a few.
Golden rice
Golden rice is a genetically modified...
Golden rice
Golden rice is a genetically modified...
5.5K
Next-generation Sequencing
97.7K
The first human genome sequencing project cost $2.7 billion and was declared complete in 2003, after 15 years of international cooperation and collaboration between several research teams and funding agencies. Today, with the advent of next-generation sequencing technologies, the cost and time of sequencing a human genome have dropped over 100 fold.
Next-Generation Sequencing Methods
Although all next-generation methods use different technologies, they all share a set of standard features....
Next-Generation Sequencing Methods
Although all next-generation methods use different technologies, they all share a set of standard features....
97.7K


